Skip to content

Space

Space 

Space()

Problem definition container for PATHOS search algorithms.

Declare what your problem can do using decorator hooks. The auto-solver selects the most powerful compatible algorithm.

Example

space = Space().initial("A") @space.successors ... def expand(state): yield "go_B", "B" @space.goal ... def is_goal(state): return state == "B" result = space.solver().solve()

Source code in pathos/core/space.py 
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
def __init__(self) -> None:
    self.capabilities: set[Capability] = set()
    self._initial_value: Any = None
    self._initial_factory: Callable[[], Any] | None = None
    self._timeout: float | None = None
    self._mode: Mode = "auto"
    self._cancel_token: CancelToken = CancelToken()
    # Total wall-clock deadline (perf_counter() basis) set by Solver
    # when a timeout is given. Meta-algorithms read it to allocate
    # per-phase budgets without needing to know the timeout themselves.
    self._deadline_at: float | None = None
    # Per-phase deadline installed by meta-algorithms (e.g. AnytimeLocal)
    # so cooperating phases break early via `_cancel_requested()`. None
    # means no phase-level deadline; only the cancel token / global
    # deadline apply.
    self._phase_deadline_at: float | None = None
    self._n_workers: int = 1
    self._adversarial: bool = False
    self._players: int = 2
    self._maximizing_player: int = 0

    # These are set via decorator hooks before any algorithm uses them;
    # typed as Any to avoid false-positive "None not callable" mypy errors.
    self._successors: Any = None
    self._goal: Any = None
    self._heuristic: Any = None
    self._evaluate: Any = None
    self._terminal: Any = None
    self._utility: Any = None
    self._reverse_successors: Any = None

mode 

mode(mode: Mode) -> Space

Declare the selection mode for the auto-solver.

  • "exact" (current default — flips to "auto" once AnytimeAStar is registered): admissible algorithms preferred.
  • "approximate": bounded-suboptimal A* variants outrank exact ones — useful when A*'s admissibility bill is too expensive.
  • "auto": cascade meta-algorithm wins selection (anytime delivery: best incumbent at any point in time).

Mirrors the .timeout() pattern: the value is read by Algorithm.score_for(space) at selection time.

Source code in pathos/core/space.py 
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
def mode(self, mode: Mode) -> Space:
    """Declare the selection mode for the auto-solver.

    - "exact" (current default — flips to "auto" once AnytimeAStar
      is registered): admissible algorithms preferred.
    - "approximate": bounded-suboptimal A* variants outrank exact
      ones — useful when A*'s admissibility bill is too expensive.
    - "auto": cascade meta-algorithm wins selection (anytime
      delivery: best incumbent at any point in time).

    Mirrors the .timeout() pattern: the value is read by
    Algorithm.score_for(space) at selection time.
    """
    if mode not in _VALID_MODES:
        raise ValueError(
            f"mode must be one of {sorted(_VALID_MODES)}, got {mode!r}"
        )
    self._mode = mode
    return self